This project explores how artificial intelligence interprets and understands typefaces, their anatomy, and their classifications, and how possible misunderstandings might result either from poor human “communication” or from the program finding patterns that humans don’t necessarily perceive.
The broad array of generated typographical artefacts is re-categorised, re-arranged, and their underlying shapes — the “skeleton” of the generated letterforms — are ultimately used in the design of a typeface with eight variations, each reflecting an emerging “bias” of the model.